006 - Big fat failure
Hello, and welcome to Genetics Unzipped - the Genetics Society podcast with me, Dr Kat Arney. In this episode we’re looking at the genetics of failure - why we fail to lose weight thanks to our genes, and why ignoring genetic information and DNA diversity leads to billions of dollars being wasted on drugs that don’t work.
Skinny genes
Kat: It’s the middle of February, which means that for many people, New Year’s Resolutions are probably starting to slide. But if you’re struggling to stick to salad the explanation might lie at least partly in your genes, according to recent research showing that people who manage to stay skinny seem to have particular genetic variations compared to those of us who are on the plumper side - or, as I like to think of myself, built for comfort not for speed.
To find out more about the curious connections between our genes and our waistlines, Graihagh Jackson went to chat with Dr Giles Yeo from the University of Cambridge who was at the Royal Institution in London last month to launch his new book - Gene Eating: The science of obesity and the truth about diets. He started by explaining why weight loss is just physics, but the process of losing, gaining or maintaining weight is all about our biology
Giles: So if you have an input, output imbalance of seven calories per day, which is not a lot - two and a half Tic-Tacs or half a squeeze of ketchup, right? - You are, from the maths going to gain close to 15kg over 30 years. Now, it's slightly simplistic but the actual physics is going to be right. And it's about the physics. The physics is simple. How you gain weight and how you lose weight is simple, meaning that you have to eat more than you burn to gain weight, and to burn more than you eat therefore, to lose weight.
That's the explanation for and the solution to the obesity epidemic. The problem is: How? - which is physics - is not the right question to ask, I'm going to argue. The question is: Why? So why do people eat more in this food environment? Why do people behave differently around food? And that is where there is powerful genetic biological underpinnings.
And you might think that it's semantics. "Oh, it's how, it’s what, you're playing with words. It's got nothing to do with nothing." Well no, that's not true because if the fulcrum of the problem is around physics - How? - then a one-size fits all solution of eat less, move more should work. But it doesn't. And it doesn't because we all behave differently around food and so a solution to get us to eat less is going to have to be personalised and individual.
Graihagh: Can you tell me a bit about the -- is it the Pima Indians in Arizona? Because I think that really underpins the question of why some people tend to act differently around food, at least from a genetic perspective?
Giles: So there are certain groups of people, ethnicities so to speak, who are more likely to become obese than others. The Pima Indians are one, and the Pima Indians are not South Asian Indians, they are indigenous peoples of America, so Native Americans.
The Pimas were from around the Arizona area, near the Gila River. At some point, some of the Pimas around the Arizona area moved to the mountains of Central Mexico, and I don't know what the reasons are. But in effect you have two groups of Pima Indians who are genetically identical, or as identical as two ethnic groups would be.
Now, this move happened ages ago, but what has happened is obviously, ages ago our environment was roughly the same, but then what has happened now is that the Arizona Pimas have been exposed to the American diet and now 50 percent of them have type 2 diabetes.
Graihagh: Wow!
Giles: 50 percent of them, against a world average of 15 percent, and we are in an epidemic! Whereas, in the Pimas that actually live in a more agricultural society in the mountains of Central Mexico, they have hardly any type 2 diabetes at all.
And so people have been studying them for a long time and the bottom line is this: This has actually brought about what we call the thrifty gene hypothesis. It's slightly simplistic because it's not one gene, but there we go – it’s the thrifty phenotype hypothesis, whereby because the Pimas in Arizona lived in a very harsh environment, they had adapted to be more efficient around their food, to eat more when there was food and to store more. So for every given calorie, to store more than they burn whilst doing the same amount of work.
When the Pimas moved to Mexico that was fine but then when the environment changed in what is now the Tex-Mex area, suddenly you have this extreme response. You get this extreme response because they didn't want to move away from the Gila River because there's some spiritual connection there. But because they couldn't move from it and they were tied to the area, they had to respond or they were going to die. Whereas, if you go look at the Pimas in Mexico, they were skinny. Here's the bottom line: Genetically they are identical. The difference is within their environment.
Now, this sounds very exotic and very severe. You might think this is nothing to do with nothing, but this is actually just a more extreme description of what is actually happening to the world around us - where all of us in the past never had enough food and so we evolved in a time where when there was food we ate, because we were never guaranteed when the next meal was going to be there.
In a time of plenty, our genetics mean that we are trying to prepare ourselves for a famine that's never going to arrive. So these genes which may have been advantageous during the feast, famine cycles have now become toxic in a feast, feast environment.
Graihagh: So, how hereditary is weight? You talk about it in your book being a bit like height, and from everything you've just described here, it sounds like the environment is a really big, important factor in this obesity crisis. But you study genes, so where did the genes come into this picture?
Giles: The number you are looking for is what we call heritability. And heritability is an interesting number, because it doesn't mean the percentage of something that comes from genes, but the percentage of the variation of a given trait which has to do with genes.
This number tends to be calculated using twin studies. You have identical twins and you have non-identical twins, and identical twins share 100 percent of their genes and then non-identical twins share as much as you would with your brother or sister, so 50 percent of your genes. You take any given trait and ask, how much -- if you share 100 percent of DNA versus 50 percent of DNA, what is the variation, how much is alike? And you can calculate the heritability from there.
I'll give you just two examples, first. Hair colour is going to be very powerfully genetically influenced with very little environmental input, and bleach doesn't count for that. Whereas freckles are also very powerfully genetically influenced, but whether or not they appear, even amongst identical twins, depends on whether or not you like to wear t-shirts or you like to stand in the sun. So a genetic trait with powerful environmental influence. Weight sits somewhere in between. Around 70 percent heritability is our body shape and body size.
Graihagh: As much as 70 percent?
Giles: As much as 70 percent. But it's not 100 percent, and that's amongst identical twins, so clearly that number can shift. Also, your socio-economic class, your environment will also shift that. The range is actually anywhere between 40 and 70 percent depending on your socio-economic class, your lifestyle. So the environment can shift it, you see? There is a genetic role to be played, but the environment can influence the percentage which your genetics actually play.
Graihagh: So we shouldn't lose hope?
Giles: I don't think we should lose hope. I mean, I've got two analogies which I use. The first one is the hand of poker analogy. If we imagine your genes to be a hand of poker, and you can have good hands and you can have bad hands and you can't do anything about that. You can blame your folks, that's about the only thing you can do! But - you can win with a bad hand of poker, even though it's more difficult. In other words, it's not impossible.
Another analogy which I will use is I am Chinese by ethnicity and I will never, ever run as fast as Usain Bolt. That's down to my genes. That's what I tell my wife, anyway. But it doesn't mean that if I train, I won't be able to run faster than I do now. So what your genes do is, yes, of course your genes bracket some possibilities. You know, why am I born with my genes? Why do I look like who I am? How fast can I run? There are going to be genetic limits set on that. It's a bad thing to say, but look, it's true. But it doesn't give you a point in time.
What your genetics does is to give you a range. When we're trying to move ourselves within that range, bodyweight and our response to this environment is going to be one of those. So there's something we can do about it. It's just more difficult for some people than for others.
Graihagh: If it is 70 percent heritable, obesity or how likely you are to gain weight or not, does that mean we can predict it by looking at say, your genes or even your parents' genes?
Giles: That's an interesting question. Because you can now get these genetic tests, by 23 and Me, by DNA Fit, you can go to any drugstore or chemists and actually buy these. It costs anywhere between £100 and £250 and they claim to be able to predict any number of things. How fast you can run, your aerobic capacity and your likelihood of becoming obese, based on the fact that there is a genetic influence to bodyweight, which I do agree with.
The problem is, this is still not possible given what we know. The reason behind that is because -- these companies are not lying, I just want to point this out: this is based on good science that has been published and it's based on these genetic risk scores on a population level which shows you can have an increased risk of becoming obese or not - the problem is they are fundamentally misunderstanding the difference between population level risk and individual prediction.
I'm going to give you an example. So, if we take something like reproduction in human beings, I think we all know that the younger you are as a female, the more likely you are to get pregnant if you're trying to. And there's a curve, you can actually draw it. Billions of women have become pregnant and had babies, and the older you get, the lower the likelihood, until you get to the menopause and the likelihood is zero. It's a very, very, very robust curve and we understand the biology behind it.
Yet we cannot take a random 34-year-old woman off the street and predict if she is going to become pregnant, just based on this curve. Why? Because it could be 100 percent because she is ovulating that day. It could be zero, because she could be infertile for a myriad of different reasons.
The reason is, you need more biological information in order to make a prediction, even against the background of population level risk. And this is the same for your genetics. Your genetics, because they give you a range and because it doesn't determine who you are - it influences who you are, it doesn't determine who you are - you need more information.
So the companies are only taking the genotype, your actual genetic information from a few spots in a genome, and trying to make predictions they cannot make. At the moment, we are still unable to predict. There may come a time where we will, but not today.
Kat: That’s Giles Yeo speaking with Graihagh Jackson at the Royal Institution. Giles’ new book, Gene Eating, is out now.
#FAIL
Kat: Also last month, our intrepid reporter Martha Henriques headed off to the Festival of Genomics in London, organised by genetics news service Frontline Genomics, to find out what’s going on in the world of genes, genomes and DNA. While there was a lot of excitement about using genetic information to inform and personalise healthcare, there are still several issues that need to be addressed along the way.
Around 90 per cent of all drugs in development fail at some point along their journey from the lab to the patient, usually because they don’t work, representing a huge waste of time, money, and lives. So first of all, as Martha found out when she spoke to immunologist Paul-Peter Tak, Venture Partner at Flagship Pioneering in Cambridge, Massachusetts, there’s a lot more that could be done when it comes to using genetic data to develop more effective drugs.
Paul-Peter: So, first I think it is critical to become much better at identifying the best therapeutic targets. People have often been misled and tried to invest quite a lot to make molecules against the wrong targets. It has become very clear that while you need clinical models to test for safety and toxicology and mechanistic understanding, it's very important not only to base your research on animal models, but to learn as much as possible about human biology. I think this is where genetics and genomics can be very helpful, but it does not solve the whole problem. It's very important to do more than that.
Martha: How far can better understanding of the human genome help to reduce rates of attrition?
Paul-Peter: Of course, time will tell with better techniques and better approaches, it may increase. But there has been a paper showing that you can double the probability of the success of a medicine in early discovery if you have genetic validation. Whatever way you look at it, the probability of success goes up from 3 percent to 6 percent.
That's amazing when you have twice as much success, but it still means that there is 94 percent uncertainty. So I think it clearly illustrates that you need a much more holistic approach to reduce failure, in particular in late stage development, where the big money kicks in and when it becomes very expensive to do large, regulated clinical trials.
Painting a protein atlas
But there’s another issue - and that’s the tricksy fact that genes are merely recipes for making the hundreds of thousands of different molecules that make up the cells of our bodies, known as proteins. But just as there’s a lot of room for error, alterations and improvisation between a neatly typed recipe in a book and the resulting cake coming out of your oven, there is still a pretty big black box between your genes (or genotype) and the actual molecular makeup and behaviour of cells, tissues and organs (that’s the phenotype).
Each gene can make many different proteins, and different types of cells will need different assortments of proteins to function properly. So to really understand what’s going on in a healthy person’s body, we need to know exactly which proteins their cells are making, as well as just the instructions in their DNA. And it’s these protein molecules that most drugs are designed to target, rather than genes themselves.
Mapping all the proteins in human cells is a much harder task than you might imagine - certainly far harder than just getting hold of a genome - as Martha discovered when she spoke to Cecilia Lindskog from Uppsala University in Sweden who is Director of the Tissue Atlas, part of the the Human Protein Atlas - a massive project aiming to answer exactly this question, by creating antibodies that specifically recognise all the various proteins produced by human cells, then looking to see where they are.
Cecilia: So we select a region of each protein that is specific and unique so we know that this antibody is only targeting this protein that we generate. And when we use the antibodies with a method called immunohistochemistry, this generates a staining reaction, so you can visualise where the antibody has bound to the protein because then we will get a brown colour, and we can look at it in microscopic images manually and study the proteins in the cells by looking at them.
Martha: So with your human samples that you are using, are these mainly from healthy patients or diseased patients or a mixture?
Cecilia: We have samples from more than 40 different normal organs throughout the entire body and also 20 different types of cancer. So we are covering most of the normal organs and the most common cancer types.
Martha: With the information that you are generating with the Atlas - who do you think that's going to be useful for and what potential treatments or clinical developments do you foresee coming from this?
Cecilia: We hope that the data provides the basis for further research in different organs. As we provide information, how the proteins should be expressed under normal conditions and then that can be compared in different studies in different diseases.
But we also believe that the information should be very relevant for pharma companies, for example, when developing new drugs - knowing where the proteins that the drugs will be targeting are located throughout the human body - because then you could potentially learn more about the future side effects of the drug and so on.
Martha: And so presumably, this is quite a large and long timescale project, as humans have more than 20,000 genes. So looking at all of those gene products must be taking you some time, I imagine?
Cecilia: Yes! We've been doing this project since 2003. We are generating antibodies towards each of these proteins and look through hundreds of images for each protein, so it has been taking us more than 15 years. But now we have coverage of more than 80 percent of all the human proteins that we have mapped with antibodies.
Martha: You say you are about 80 percent coverage. Is it getting harder as you go through to the last proteins? Are there any that are particularly difficult to work with?
Cecilia: Yes, it is. We have actually been at 80 percent for the last four, five years. It's difficult to increase to 100 percent because some of these proteins may only be expressed under certain conditions or certain functional circumstances. Or, they are expressed in very specific tissues that we didn't look into previously, or maybe only during foetal development and not in adult humans. This is also a level where you could go more into details to try to cover these last percentages. Or they are proteins that are very difficult to map with antibodies because they might be secreted and are not found in the tissue or so on.
Martha: There are presumably a finite number of human proteins - do you see that this project will ever come to an end? Will you ever conclusively say, "Yes, we have finished the human protein atlas."?
Cecilia: It depends on which level you want to do it, because you can always continue and dig deeper into more details. We can start to add more specific tissue types that we didn't look into previously. They are not the most common tissue types, like, more specific regions of the brain, ear, eye and so on.
So I definitely believe that the project could go on for decades if we want to go into more detail. We hope that this project can provide the basis for further research around proteins. This is the first draft, but definitely more can be done at different levels.
Martha: Do you have an idea of when you may approach that eventual 100 percent mark?
Cecilia: We are aiming for increasing the coverage. So we're hoping to have 90 percent within the next three or four years.
Martha: But it's hard to put a number on when you'll be finished?
Cecilia: Yes, it is!
Kat: Cecilia Lindskog, who appears to have embarked on the molecular biology equivalent of painting the Forth Bridge.
From human genome to global genome
Kat: So - in order to make more effective treatments and tests, not only do we need to use genetic data to inform drug development, we also need to know more about how genetic makeup actually relates to the proteins present in a cell. Then there’s another angle: genetic diversity.
The vast majority of all the data in current genomic databases that are used to inform the development of new tests and treatments comes from people with white European ancestry, which doesn’t accurately reflect the genetic makeup of the global population. This means that billions of people around the world stand to miss out on future medical advances that might not work for them or could even be harmful.
To find out more about the implications of the lack of genomic data diversity, Martha spoke to Paul Matthews, head of strategic partnerships at Global Gene Corp - a company that I’m proud to be working with - who are gathering and analysing genetic data from all around the world, particularly India and Africa.
Paul: This is a very big issue at the moment. There's a publication and a review of GWAS studies only two weeks ago, which highlights what we've found within our company over the years, that there's a big skew towards European ancestry, Caucasian data. And it has been right from the very start, when the first Genome project was completed and subsequently.
The reason being for the bias that the funders and actually where the research is being done has been in the US and the UK and in Europe. So there's been very little real data coming from other ethnicities and parts of the world. 81 percent of all of the data was Caucasian.
If we look at, say, India, where we're in operation, it has 1.3 billion people, 20 percent of the global population, yet represents less than 1 percent of the global data.
Martha: So even within typically Caucasian countries such as the US and the UK, we are actually a very diverse country - is there also an issue with access to genetic sequencing that is implicit in people's ethnic backgrounds, as well?
Paul: Well, I'm not so sure on that front. I think one of the issues has been actually, who is being sequenced and why? We know that this year, or at the end of last year in October, the NHS have said that they will sequence particular groups of patients in their genomic medicine service. I think also, listening to one of the speakers here yesterday who was suggesting that traditionally, some of the minority ethnicities have viewed sequencing with suspicion. That may be due to a lack of materials translated into a language which they're familiar with.
There's no one easy answer to why that's happening, but given that the vast majority of data is white Caucasian ethnicities, it's just reflective of the demographic in those different countries, not just linked to healthcare but also to research and the funding. So if funders are funding in a particular country, they are focusing in on the demographic of that country, and to actually fund work outside, say in India or Asia or Africa, is very difficult.
Martha: So what would a perfectly diverse genetic database look like? In what ways would it have to be diverse?
Paul: Again, a very good question. We know that there's a real skew to Caucasians. If we look at the Asian data, a lot of the Chinese data is in a situation where it can't be used by the community, it's kind of locked away. We are trying to expand the available Indian data.
There's very little in Africa, but there are some good initiatives that are trying to redress that. I think again, funding is an issue there. There's very little South American data in the databases and in the Middle East there are various sequencing programmes going on, but again it's early days.
Martha: What problems arise as a result of a lack of diversity in genetic databases?
Paul: Well, we have this idea of the human genome. We have a reference genome. But the reference genome is very much biased towards a European ancestry reference. So for example, if we look at different allelic architecture studies, certainly from the work and the research that we've done with our data in India, we see very different genetic makeups to other populations.
Coming back to the idea of what would make a truly diverse dataset, if we look into India, there are 4,500 to 5,000 different ethnic groups, subpopulations across India itself. So it's going to be very difficult to say we can produce a reference Indian genome or a reference African genome or a reference South American genome. It's a very broad area, and without data coming from many parts, it would be difficult to say that we have solved the problem.
Martha: Another problem that's been raised is around the use of pharmacogenetics in the NHS after its long-term strategy has recently been announced. So what issues would arise there, if there was a lack of diversity in the databases being used?
Paul: Well again, pharmacogenomics is looking at how specific variants interact with drugs and drug-gene interactions. So if we don't have a true representation of different ethnicities and how particular variants operate within those groups, it would be very difficult to generalise, to say that we have a pharmacogenomics panel, or a pharmacogenomics set that is appropriate to everyone - because it won't be appropriate to everyone, it will be appropriate to those people whose genomics have been used in the first place.
For example, if we look at some of the studies in India, if we separate out two states, we've looked at data for a commonly prescribed immunosuppressant drug. If we look at one state, there's a pharmacogenomic variant which will determine the response to that drug being prescribed.
In one state, 60 percent of the population have one particular allelic variant and on the other side of the continent there's a state where they have an 11 percent frequency of that allele. So, prescribing that drug within the two different states in the same country gives a very different picture.
Martha: How optimistic are you? At the moment, do you think we are moving in the right direction or do you think things are going to get worse before they get better?
Paul: I think we are definitely moving in the right direction. The recognition that diversity needs to be enhanced is a good one. I think there could be further dialogue between companies like ourselves and funders and health services to see what data is available, what data is needed. And I think that there are collaborations that can be done proactively from both sides.
Kat: Global Gene Corp’s Paul Matthew speaking with Martha Henriques, rounding off our report from the Festival of Genomics, and thanks to the team at Frontline Genomics for the invitation to the festival.
Evolution’s gland plan
And finally, think of Charles Darwin and you’ll probably recall his famous voyage on the Beagle - possibly the greatest gap year in scientific history - where he made his observations of exotic animals in distant lands that helped to shape his groundbreaking theory of evolution by natural selection. But Darwin’s ideas actually owe much more to his studies of the domestic chickens in his backyard than the glamorous Galapagos finches, which were also subject to selection (albeit by human hands rather than natural forces).
In the latest podcast from Heredity, The Genetics Society’s journal, James Burgon chats to Dr Amir Fallahshahroudi about his recent work investigating the genetic basis of domestication in the chicken, which has helped to turn jumpy wild junglefowl into today’s farmyard chooks.
James: In the study, you zoomed in on gene expression in the pituitary gland - why is it that the pituitary gland is interesting in terms of domestication?
Amir: The pituitary gland lies between the brain and a lot of different glands in the body. So it gets a signal from the hypothalamus, and also, it's going to give the signal or is directly going to affect other organs in the body. The surprising thing in this study was almost everything was within a dream of our expectation. I was not expecting to find this many key genes differentially expressed in the pituitary gland.
Kat: You can get the full interview in the latest Heredity podcast. For more information about this podcast including show notes, transcripts, links, references and everything else head over to geneticsunzipped.com You can find us on Twitter @geneticsunzip or email us at podcast@geneticsunzipped.com with any questions and feedback. Please do take a minute to subscribe on Apple Podcasts or wherever you get your podcasts from, and it would be great if you could rate and review - and more importantly, please spread the word. Tell your friends, send out a tweet, and share the love.
In the next episode we’ll be exploring more of our 100 ideas in genetics by casting an eye over the careers of some of the world’s top models - model organisms, that is...
Genetics Unzipped is presented by Kat Arney and produced by First Create the Media for the Genetics Society - one of the oldest learned societies in the world dedicated to supporting and promoting the research, teaching and application of genetics. You can find out more and apply to join at genetics.org.uk Our theme music was composed by Dan Pollard, and the logo was designed by James Mayall. Thanks to our reporters Graihagh Jackson and Martha Henriques, thanks to Hannah Varrall for production, thank you for listening, and until next time, goodbye.
Transcription by Viv Andrews
References and links:
Gene Eating: The science of obesity and the truth about diets - Giles Yeo
Giles Yeo - Twitter
Cecilia Lindskog - Twitter, ResearchGate
In it to win it: addressing the lack of global genomic data diversity - Global Gene Corp
Global Gene Corp - Twitter
